CAM Row Hammer Refresh with Multi-Phase Address Capture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional row hammer refresh methods for content-addressable memory devices suffer from low row hammer address detection due to lengthy 'no sampling periods' caused by storing only the first few addresses received during a refresh interval, leading to inefficient data refresh and potential data corruption in adjacent memory cells.
Innovation Solution
Divide the row hammer refresh interval into multiple phases, randomly capturing and storing unique row hammer addresses in separate registers during each phase, and refreshing them consecutively at the end of the interval, thereby enhancing detection and reducing 'no sampling periods'.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If only the first few addresses received during a refresh interval are stored, then the device complexity is reduced, but the row hammer address detection precision deteriorates
Solution Approach 1:
The patent divides the address storage function into multiple independent registers (first register, second register, third register, fourth register), each capable of storing a unique row hammer address. This segmentation allows simultaneous storage of multiple addresses without increasing overall system complexity, as each register operates independently during different phases of the refresh interval.
Solution Approach 2:
The patent introduces a temporal dimension by dividing the refresh interval into multiple phases, with each phase associated with a specific register. Addresses are captured in different time phases and stored in corresponding registers, transforming a single-dimension storage problem into a multi-dimensional solution that improves detection precision without proportional complexity increase.
2Measurement precision
If multiple unique row hammer addresses are captured and stored in separate registers during different phases, then the row hammer address detection precision is improved, but the device complexity increases
Solution Approach 1:
The patent implements periodic address capture by dividing the refresh interval into repeating phases, where each phase periodically captures addresses and stores them in corresponding registers. This periodic structure allows systematic monitoring of multiple addresses over time while maintaining a regular, predictable pattern that simplifies control logic despite the increased number of registers.
Solution Approach 2:
The patent performs preliminary address capture and storage in separate registers during different phases before executing the refresh operation. By pre-capturing and organizing addresses in dedicated registers during phases 1-4, the system prepares the necessary data structure in advance, enabling efficient subsequent refresh operations without requiring complex real-time decision-making.
3Productivity
If the refresh interval is divided into multiple phases with random address capturing, then the productivity of address detection is improved, but the ease of operation deteriorates
Solution Approach 1:
The patent introduces dynamic address capturing by randomly selecting addresses during each phase rather than following a fixed sequence. This dynamic approach improves detection productivity by capturing actual row hammer addresses that occur during normal operation, while the underlying phase structure and register assignment remain static and predictable, maintaining operational simplicity.
4Reliability
If consecutive refresh operations are performed for multiple stored addresses, then the reliability of data protection is improved, but the loss of time increases
Solution Approach 1:
The patent performs preliminary address capture and organization in separate registers during phases 1-4 before executing the refresh operations. By preparing all address information in advance and organizing it in dedicated registers, the subsequent consecutive refresh operations can proceed efficiently without interruption or additional decision-making, minimizing the actual time penalty despite the increased number of refresh cycles.
Data Source
AI summary
A method of operating a memory device may include receiving, during a phase of a row hammer refresh (RHR) interval, at least one row hammer address (RHA) of a content-addressable memory (CAM). The method further includes storing, during the phase of the RHR interval, a received RHA of the at least one received RHA in an address register. Further, the method includes refreshing the stored RHA of the CAM via a RHR during the RHR interval.